1. Introduction
Perceived lifetime discrimination has been identified as a critical set of experiences that shape different health profiles between individuals reporting different levels of discrimination (Bound et al., Reference Bound, Geronimus, Rodriguez and Waidmann2015, 2000; Goosby et al., Reference Goosby, Cheadle and Mitchell2018; Lacy, Reference Lacy2007; Pattillo, Reference Pattillo2013). Research shows that the differential stigma and discrimination that individuals experience throughout their lifetimes are patterned across socioeconomic strata as well as gender and racial lines in the United States and abroad (Echeverria-Estrada et al., Reference Echeverria-Estrada, Javier and Claudia2020). In the United States, different levels of structural discrimination are distributed across sociodemographic groups through the political apparatus—e.g., from the housing policies that have historically created residential segregation and the criminal justice system that creates mass incarceration to the immigration policies that separated immigrant families and held their children in detention camps. As residents of the United States go on to live their daily lives, they interact with historically embedded, now institutionalized features of society that expose them to racialized experiences of privilege and discrimination.
Fundamental Cause Theory (FCT) research has shown that socioeconomic inequalities differentially distribute health outcomes across demographic groups, and that such differences can persist over time. Individuals with a high socioeconomic status tend to use their resources—e.g., power, knowledge, money, and prestige—to avoid risks and to protect themselves. Recent studies in FCT address how these processes result in long-term social stratification, stigma and discrimination, and structural racism and how these, in turn, affect health outcomes—e.g., by affecting people’s chances to access timely and high-quality medical treatments (Dorling, Reference Dorling2015; Rodriguez et al., Reference Rodriguez, Bae, Geronimus and Bound2021; Ye and Rodriguez, Reference Ye and Rodriguez2021).
Research has also shown that the health status of individuals depends on where they locate their life experiences in the privilege-discrimination spectrum (Assari and Caldwell, Reference Assari and Caldwell2018; Blackwell et al., Reference Blackwell, Iacus, King and Porro2009). Unfortunately, the sociopolitical conditions that guide the lifetime discriminatory experiences that limit the personal and socioeconomic development of racial and ethnic communities, are entrenched (Rodriguez, Reference Rodriguez2019). Detrimental features of U.S. society, like structural and institutional racism, remain stubbornly correlated with racial disparities in health, in spite of powerful governmental and non-governmental interventions. The strong correlation between perceived discrimination and health status is also robust across time and space. In the United States, this correlation can be observed across populations with different disease and health risk profiles, different cultural and demographic characteristics, and across different as well as political and economic settings. Institutional racism has proven to be extremely flexible, adapting and re-adapting itself to public and private efforts designed to end it.
Understanding differences in perceived lifetime discrimination by race and socioeconomic status (SES) is critical for our understanding of many stress-mediated social determinants of health. This is because perceived lifetime discrimination captures some of the critical structural and individual variation related to the processes that distribute stress-related physiological dysregulation throughout the lifespan (Rodriguez, Reference Rodriguez2018). The perception of unfair treatment is essential given that the individual is the ultimate recipient unit of discriminatory processes—irrespective if the process originated at the structural or individual level, or if it is explicit or implicit. Individual-level perceptions of discrimination clustered by race and SES inform us about the structural paths that thread the social fabric and the public policies to address them. No serious policy model for shrinking racial and SES disparities in health is conceivable without incorporating a rigorous strategy to drastically reduce lifetime discrimination on the basis of class and race.
The study of lifetime discrimination gains additional salience as the types of discrimination analyzed in this study have the potential to thwart lifetime socioeconomic opportunities to racial and ethnic minorities. As these discriminatory experiences inhibit access to social, economic, and political resources, they have been found to strongly correlate with lower personal development, and worse health and wellbeing (Cardarelli et al., Reference Cardarelli, Cardarelli and Luz Chiapa2007). These associations suggest that, as race and SES are confounded, a different profile of unexplored outcomes may arise from their interactions, including in experiences of discrimination and health. For instance, research has shown that key social determinant of health like education and income bring weaker health benefits for Black people than for whites (Assari and Caldwell, Reference Assari and Caldwell2018; Blackwell et al., Reference Blackwell, Iacus, King and Porro2009). Such non-equivalence patterns of SES protective effects for Black people are also manifested by individuals with other characteristics of vulnerability (Assari and Lankarani, Reference Assari and Lankarani2016; Hudson et al., Reference Hudson, Bullard, Neighbors, Geronimus, Yang and Jackson2013).
Such diminished returns to SES resources among Black people bring to light some overlooked yet critical policy complexities, especially those anti-discriminatory efforts aiming at reducing health disparities. Policies and programs that improve income and education can also be applied to improve health outcomes in vulnerable communities. There is urgency for designing anti-discriminatory policies that would improve overall health as well as racial health inequities. From the perspective of reducing racial disparities in health, desirable improvements in education and income are imperative to improve the health of all, but they may also be over-simplistic as they can also increase health disparities between racial groups (Assari and Caldwell, Reference Assari and Caldwell2018; Blackwell et al., Reference Blackwell, Iacus, King and Porro2009). These diminished returns of education and income are observed to increase differences across a great variety of health outcomes and sociodemographic groups. For instance, although increasing income and education improved depression, sleep latency, hypertension, suicide, and mortality outcomes, such improvements also showed to be higher for whites than for Black people in the U.S. These results are also detected in a risk factors such as obesity, drinking, smoking, eating a healthy diet, and exercising (Ayalon and Gum, Reference Ayalon and Gum2011; Gee and Walsemann, Reference Gee and Walsemann2009; Lewis et al., Reference Lewis, Everson-Rose, Powell, Matthews, Brown and Karavolos2006) across all age groups (Beatty Moody et al., Reference Beatty Moody, Taylor, Leibel, Al-Najjar, Katzel, Davatzikos, Gullapalli, Seliger, Kouo, Erus, Rosenberger, Evans, Zonderman and Waldstein2019; Cuevas et al., Reference Cuevas, Ong, Carvalho, Ho, Chan and Allen2020).
2. Diminished Returns, Health, and Discrimination
In a study, although racial disparities in low birthweight were significant in the context of maternal age and neighborhood socioeconomic characteristics, the greatest disparities were found between African-American and Caucasian adolescents that lived in areas of higher socioeconomic status (Watson et al., Reference Watson, Otey, Westbrook, Gardner, Lamb, Corrigan and Fenton2004). In another example, although higher SES showed a protective effect for whites in Milwaukee, it did not have the same protective effect for Black people (Hankivsky et al., Reference Hankivsky, Grace, Hunting, Giesbrecht, Fridkin and Rudrum2014). Some possible solutions to such observed SES diminished returns have been proposed. Some of these solutions focus on equalizing work conditions and benefits as they correlate with educational attainment. For example, benefit-equalizing interventions at the industry level, or in non-profit or government institutions, should consider that lower education correlates with worse overall working conditions, lesser control and decision-making independence, lower salaries and time-off (e.g., vacation and sick leave), and that these disadvantages are more intense for Black people compared to White people (Lewis et al., Reference Lewis, Everson-Rose, Powell, Matthews, Brown and Karavolos2006). Other strategies are psychosocial, including socioeconomic and residential interventions through the welfare system to reduce the long-term effects of living in poverty during childhood (Austin, Reference Austin2011). Another type of solution is structural as, for example, non-poor Blacks are more likely to continue living in poor neighborhoods (Barry, Reference Barry2014) than their White counterparts (Haas et al., Reference Haas, Krueger and Rohlfsen2012) and, therefore, the improvement of poor neighborhood infrastructure may improve the returns of SES for Black people (Barry, Reference Barry2014).
Another set of studies use a biopsychosocial approach to understand how discriminatory experiences affect the intersections between the environmental context, health, SES, and race. Findings of this research show that high-achieving Black people—i.e., highly educated, high-earning individuals—report higher levels of discrimination in a variety of social contexts. When looking at how these stressful experiences correlate with health outcomes, these studies persistently show Black people at a higher disadvantage relative to their White counterparts (Haas et al., Reference Haas, Krueger and Rohlfsen2012). A long list of detrimental stress-related health outcomes has been found to correlate with the perception of discrimination (Mustillo et al., Reference Mustillo, Krieger, Gunderson, Sidney, Mccreath and Kiefe2004) including cardiovascular disease (Mustillo et al., Reference Mustillo, Krieger, Gunderson, Sidney, Mccreath and Kiefe2004; Nazroo, Reference Nazroo2003), blood pressure (Nazroo, Reference Nazroo2003; Rodriguez et al., Reference Rodriguez, Karlamangla, Gruenewald, Miller-Martinez, Merkin and Seeman2019) kidney disease (Szanton et al., Reference Szanton, Rifkind, Mohanty, Miller, Thorpe and Nagababu2012), low birthweight (Coley et al., Reference Coley, Nichols, Rulison, Aronson, Brown-Jeffy and Sd2015)and self-reported health, physical, and mental health (Goosby et al., Reference Goosby, Malone, Richardson, Cheadle and Williams2015; Priest et al., Reference Priest, Paradies, Stewart and Luke2011; Sidanius et al., Reference Sidanius, Peña and Sawyer2001). Research has shown that some of the effects of discrimination are mediated through oxidative stress (Dorling, Reference Dorling2015), dysregulation in cortisol production (Lewis et al., Reference Lewis, Everson-Rose, Powell, Matthews, Brown and Karavolos2006), and inflammation (Coley and Nichols, Reference Coley and Nichols2016).
Using data from the longitudinal Coronary Artery Risk Development in Young Adults study (CARDIA), Borrell et al. (2006) assessed the health effects of perceived racial discrimination among Black people on a variety of social contexts and activities, from experiences at school and trying to get a job to trying to get a house and accessing medical care. Borrell et al. also looked at socialization dynamics within labor environments and in public settings. They find that racial discrimination was more commonly detected in men than in women and in individuals with higher educational levels, who, at the same time, reported worse physical and mental health profiles (Coley and Nichols, Reference Coley and Nichols2016).
Research shows that detrimental health effects manifest through the process of trying to get hired or through the labor dynamics that assign promotions at work (Geronimus et al., Reference Geronimus, Bound, Waidmann, Rodriguez and Timpe2019; Harrison, Reference Harrison2013; Roberts et al., Reference Roberts, Swanson and Murphy2004). Discrimination experienced in work settings contribute to diminished psychological well-being. This type of health harm is mediated by stress and it is subsequently manifested in contained economic advancement (Hankivsky et al., Reference Hankivsky, Grace, Hunting, Giesbrecht, Fridkin and Rudrum2014). Given that stress-mediated health effects and socioeconomic success go hand-in-hand, racial discrimination is ultimately reflected in the aggregate inequities observed in society which, in turn, fix racial minorities in a labor opportunity market that keeps them at a disadvantage (Geronimus et al., Reference Geronimus, Bound, Waidmann, Rodriguez and Timpe2019).
That the health of racial minorities is conditional on perceived discrimination is a finding reported in studies that analyze this association across a variety of geographical contexts. For example, some research identifies associations between racial discrimination and health in New Zealand and report that experiences of ethnically motivated attacks (physical and verbal) or unfair treatment are associated with poor self-rated health, lower physical functioning, lower mental health, smoking behavior, and cardiovascular disease (Assari and Caldwell, Reference Assari and Caldwell2018; Harris et al., Reference Harris, Tobias, Jeffreys, Waldegrave, Karlsen and Nazroo2006). Because discrimination is linked to space, other areas of research highlight the role of residential segregation. Discrimination in neighborhoods and the collective marginalization that places and maintains vulnerable groups in poor areas adversely affect health through restricted access to health and public services. These mechanisms also expose segregated individuals to social stressors such as higher crime rates, lack of leisure infrastructure like parks and walking zones, high pollution, and higher unemployment among many others (Feagin and Sikes, Reference Feagin and Sikes1994; Kim and Fredriksen-Goldsen, Reference Kim and Fredriksen-Goldsen2017). Given the connections between segregation and harmful health exposures, a recent study reported that during the COVID-19 pandemic, U.S. counties with high immigrant density and high residential segregation showed much higher infection rates than those with low segregation and low immigrant density (Gonzales et al., Reference Gonzales, Lee and Marchiondo2019).
Another branch of health research has more specifically investigated the interactive associations of race, education, and income with lifetime discrimination (Cottrell et al., Reference Cottrell, Herron, Rodriguez and Smith2019; Gee and Walsemann, Reference Gee and Walsemann2009; Hudson et al., Reference Hudson, Bullard, Neighbors, Geronimus, Yang and Jackson2013; Szanton et al., Reference Szanton, Rifkind, Mohanty, Miller, Thorpe and Nagababu2012; Ye and Rodriguez, Reference Ye and Rodriguez2021). Racial discrimination is shown to be associated with increased risk of depression among Black men with higher income and levels of education (Fuller-Rowell et al., Reference Fuller-Rowell, Curtis, El-Sheikh, Duke, Ryff and Zgierska2017). Discrimination has also been found to affect health behaviors among racial minorities, for example, increasing drinking among Black and Hispanic men (Ye and Rodriguez, Reference Ye and Rodriguez2021). Further, the relationship between education and perceived discrimination has been shown to be moderated by race even among healthy individuals, especially women (Cottrell et al., Reference Cottrell, Herron, Rodriguez and Smith2019). In examining the association between discrimination and critical biomarkers, Van Dyke et al. (Reference Dyke, Vaccarino, Dunbar, Pemu, Gibbons, Quyyumi and Lewis2017) [70] found that discrimination across SES indicators is an important stressor altering the C-reactive protein (CRP)—a biomarker of inflammatory responses. Van Dyke et al. find particularly harmful effects among Black people with high levels of education.
Perceived lifetime discrimination is multi-faceted and affects the health of vulnerable communities differently depending on their socioeconomic position. However, little is known about the different types of perceived lifetime discrimination and the effect on racial minorities with varying incomes and educational standing. Understanding how discrimination affects long-term personal and socioeconomic development of individuals is imperative to implement adequate solutions and policies for fair treatment of all races. The study of a cumulative and independent set of perceived discriminatory experiences is also valuable to better map how Black people’s discriminatory experiences have evolved through history, and how these continue to affect them today.
3. Research Questions
This study contributes to the existing literature by answering the following three research questions:
-
i. Are race-education and race-income intersections associated with an overall index of perception of lifetime discriminatory experiences?
-
ii. Have these associations changed in time and, if so, how?
-
iii. Are these race-education and race-income intersections differently associated with individual types of perceived lifetime discrimination?
4. Data and methods
The data are from two waves of the Midlife Development in the United States (MIDUS 1, 1995-1996, and MIDUS Refresher, 2011-2014). Although the MIDUS 1 study was fielded in 1995/1996, it already includes a large battery of lifetime discrimination items that were reproduced in MIDUS Refresher and, in both instances, they varied enough to capture perceptions of discrimination across a wide range of contexts and areas of human development. By analyzing the MIDUS Refresher (2011-2014) data, we are able to investigate if the findings from MIDUS 1 have persisted about two decades later. MIDUS 1 comprises a national sample of 7,108 participants, of which 5,728 responded to the questions of the lifetime discrimination battery. After all methodological and data-cleaning procedures, our final analytic sample includes 5,036 individuals.
In parallel to MIDUS 1, MIDUS Refresher was designed to replenish the MIDUS 1 sample, and it thus corresponds to a more recent dataset for the assessment of the biopsychosocial processes represented in MIDUS 1 (Brown et al., Reference Brown, Williams, Jackson, Neighbors, Torres, Sellers and Brown2000), [71]. Accordingly, MIDUS Refresher recruited a national probability sample of 3,577 individuals, of which 2,163 were included in our final analytic sample after methodological and data-cleaning procedures were carried out. The lifetime discrimination questions used in both MIDUS 1 and MIDUS Refresher asked participants if: 1) you were discouraged from seeking high education, 2) you were not hired for a job, 3) you were not given a job promotion, 4) you were prevented from renting or buying a home, 5) you were hassled by the police, 6) you were denied a bank loan, 7) you were denied or provided inferior medical care, and 8) you were denied or provided other inferior services [74].
All discrimination subscales were coded as 1=never, 2=rarely, and 3=sometimes or often (Bound et al., Reference Bound, Geronimus, Rodriguez and Waidmann2014). In addition to these eight variables—treated simultaneously as our dependent variables to answer research question 3—we also generated a sample-specific overall index of discrimination to answer research questions 1 and 2. This index of discrimination was calculated as the sum of the eight subscales listed above. Our independent variables of interest are race (1=Black, 0=White (for our race variable, we were unable to distinguish Hispanics from non-Hispanics as the MIDUS study does not include information about ethnicity), education (1=less than high school or high school, 2=some college, and 3=at least bachelor’s), and household income tertiles (1=bottom tertile, 2=middle tertile, 3=top tertile). We further controlled for age, gender, and body mass index (BMI) given that these individual characteristics are known to confound the association between SES and discrimination.
To increase the comparability between the two groups of interest—Black and White people—and between the two periods—1995-1996 and 2011-2014—we implemented a weighting procedure comprised of propensity scores matching, adjustment for missingness, and sample-specific weights accounting for survey design and post-stratification. To do this, we first applied the same propensity scores model and matching technique to both samples, separately. The propensity score—in our case, the probability of being a Black person—was estimated using a logistic regression (Clouston and Link, Reference Clouston and Link2021). Subsequently, we used an Automated Coarsened Exact Matching (CEM) technique to match Black people to their similar White counterparts on propensity score. By using CEM, we matched Black individuals to White individuals situated in the same optimized propensity score stratum (Assari et al., Reference Assari, Lankarani and Caldwell2016). To further control for residual variation associated to our estimates of interest, our main models included the same variables we used in the propensity score models, which is standard in the literature.
We also applied inverse probability weighting to account for possible bias due to missingness in our analytic samples. For each of our analytic samples, we used a logistic regression to estimate the probability that a given individual would be missing conditional on the covariates included in our main model. Also, both MIDUS 1 and MIDUS Refresher samples, were weighted to account for survey design and post-stratification. Both samples were, accordingly, made representative of the Current Population Survey (CPS)—one of the main intercensal surveys in the U.S.—counts at the moment of measurement (i.e., 1995 and 2011, respectively). For each of the samples, our final compound sample-specific analytic weights were computed as the multiplication of the three weights mentioned above. In this manner, our analyses allow us to compare within-sample Black to White individuals, and compare our associations in our two points in time (as they are representative of their respective CPS sample in the year of measurement).
To answer our first research question, we used a Generalized Linear Model (GLM) approach, with the overall index of lifetime discrimination (one per sample) as the dependent variable. After combining all the types of discrimination into a single index scale (as is customary in existing literature), the distributional shape of our dependent variable mapped into a gamma distribution, which was determined via a Park’s Test. We used a logarithmic link function and, to account for the fact that some MIDUS participants belong to the same family, we clustered our standard errors at the family level. Our GLM were estimated using Ordinary Least Squares and robust estimation. To answer our second research question, we ran our GLM regressions for each time point, separately—i.e., 1995-1996 (MIDUS 1) and 2011-2014 (MIDUS Refresher).
To answer our third research question, we used a Generalized Structural Equation Model (GSEM) approach, with all eight dependent variables included in a single model—i.e., instead of estimating equations independently, the GSEM allows to simultaneously estimate parameters in all equations to optimize the multiple-equations model fitting to the data. To account for the skewness and the count nature of our variables, our GSEM regressions applied a Poisson family distribution with a logarithmic link function. Our GSEM models were estimated using maximum likelihood estimation. Both GSEM and GLM models were weighted using the resulting analytic weights described above. The Appendix contains a formal description of our models.
5. Results
Table 1 lists summary statistics for our analytic samples—unweighted for the overall sample, and before and after matching on propensity score, by race. In MIDUS 1 (M1), the eight types of discrimination varied from an average of 0.03 (prevented from renting or buying a home and denied or provided inferior medical care) to an average of 0.22 (not hired for a job) whereas in MIDUS Refresher (MR) they varied from 0.15 and 0.16 (prevented from renting or buying a home and denied or provided inferior medical care, respectively) to 0.39 (not hired for a job). These results show that, although discrimination values have increased in time, the types of discrimination with the lowest and highest values have not changed. Reflecting this increase, the overall index of discrimination went from 0.77 in M1 to 1.90 in MR. Also, as noted by the literature, these experiences varied broadly by race. For example, the overall index of discrimination in M1 was 0.13 for Whites and 0.28 for Black people, whereas in MR it was 0.24 for Whites and 0.46 for Black people. Results show, therefore, that Black people report, on average, about double the discrimination reported by White people in both time points.
Table 1. Summary statistics of the analytic samples
Note: This table shows results for mean differences across all variables used in the analyses before and after matching. The procedure was a propensity scores matching, where the propensity scores were estimated using a logistic regression and the matching technique was Automated Coarsened Exact Matching (see text). The resulting weights are for the subsample of Whites whereas variables of the subsample of Black people are used as the calibrating distributions.
The M1 and MR analytic samples had an average age of 47 and 52 years, respectively, and both had a slightly higher fraction of females (52%). The average income of households located at the bottom, middle, and top tertiles were $27,529, $73,420 and $186,358 in M1, and $23,155, $72,222, and $160,693 in MR, respectively—which is higher than those of the overall population. Similarly, in M1 69% of the analytic sample had an educational level of at least some college whereas in MR it was 54%. The average body mass index was 26.6 in M1 and 28.9 in MR, suggesting a healthy overall sample of individuals in the two time points.
Over 99% of the original sample was successfully matched in both samples. That loss of information was minimal is critical given the relatively low percentage of Black individuals included in the sample (5% in M1 and 7% in MR). Indeed, our matching procedure diminished racial differences across covariates, with none of these differences (with the exception of income for the bottom tertile in both samples) reaching statistical significance after matching. Overall, after matching, racial differences in perceived discrimination diminished, especially in the MR sample. This contraction indicates that the factors that differ between White and Black people may be partially related to the different perceived levels of discriminatory experiences between the races, more strongly in MR relative to M1.
Both GLM and GSEM parameter estimates are reported in the Appendix. Table 2 reports the predicted values of our GLM models for education-race and income-race effects, with results for M1 at the top and results for MR at the bottom. According to these results, racial differences in overall perceived discrimination vary by level of education and income tertile. Whereas differences in overall discrimination for Whites do not differ much by education and income in both samples, they do for Black individuals. Black-White differences in overall discrimination are substantial in both samples, especially as education attainment increases. These Black-White differences reach statistical significance at conventional levels (p<0.05). The two education-race patterns are very similar between the samples, with an increase in perceived discrimination among White people in MR. A change of pattern (from M1 to MR) is more clearly depicted by income, with Black people manifesting higher levels of overall discrimination as income increases in M1 but more similar across income tertiles in MR. Figure 1 is a visualization of these effects (predicted values for all individual types of discrimination from our GSEM models are reported in the Appendix).
Table 2. Predicted values from GLM models
Figure 1. Predicted values for overall index of lifetime discrimination for the MIDUS 1 and MIDUS Refresher samples by education, income, and race.
Note: the bars denote 95% CIs. The standard errors were calculated using the delta method.
6. Discussion
We examined an overall index of lifetime discrimination, computed from eight different types of lifetime discrimination experiences, known to affect the life opportunities of individuals and stress-related processes. Our findings revealed large racial differences in perceived lifetime discrimination depending on income and educational status. Highly-educated, high-earning Black individuals reported perceiving higher discriminatory experiences than any other White-education and White-income group in the samples. White people did not only report lower levels of discrimination at all education and income levels, but, among White individuals, there were no statistically significant differences in discrimination across these levels. Similar levels of racial inequality in perceived discrimination were substantial and detected in all types of discrimination (Appendix).
Looking at our overall index of discrimination, a Black person with at least a bachelor’s degree perceived 5.4 times more lifetime discrimination than a White person of the same educational level in 1995-1996. This ratio decreased by 2011-2014, when the average Black participant in the sample reported 2.5 times more lifetime discrimination than a White person. These Black-White ratios were 3.6 and 5.2 in M1, and 1.1 and 1.8 in MR for people with a high school degree or less and people with some college education, respectively. Table 2 lists the discrimination gap between the races. Although the race-education patterns are strikingly similar in the two time points (Figure 1) and racial differences remain very high almost two decades later (Table 2), the decrease in the ratios is mostly the result of higher discrimination reported among White people (Table 2). Table 2 shows that all Black-White gaps in discrimination, at all levels of education, are statistically significant (p<0.05) in M1 but not in MR, with Black-White gaps in discrimination not reaching statistical significance for individuals of low education in MR.
This finding is important: Black people continue to see diminishing returns to education while White individuals report increasing levels of discrimination at all educational levels. Also, Whites of low-education report an increasing level of discrimination now similar to those of Black people with low levels of education. The racial gap among the least educated is shrinking, but for worse, as both groups are reporting higher levels of discrimination, with low-educated Whites’ discrimination tripling in almost two decades. Given the connections between discrimination and health, our findings are reflective of the relative deterioration of low-educated Whites, with their life expectancy declining at least since 1990 (Cottrell et al., Reference Cottrell, Herron, Rodriguez and Smith2019; Cuevas et al., Reference Cuevas, Ong, Carvalho, Ho, Chan and Allen2020; Harris et al., Reference Harris, Cormack, Tobias, Yeh, Talamaivao, Minster and Timutimu2012).
Diminishing protective returns were also large for income (Bell et al., Reference Bell, Sacks, Tobin and Thorpe2020; Borrell et al., Reference Borrell, Kiefe, Diez-Roux, Williams and Gordon-Larsen2013; Gonzales et al., Reference Gonzales, Lee and Marchiondo2019). In M1, a Black person at the top tertile of the income distribution perceived 7 times more lifetime discrimination than a White person at the same income tertile. This ratio decreased to 1.4 times almost two decades later in the MR sample (Table 2). This is the case even though the average White individual across all educational and income levels reported to perceive, at maximum, the lifetime discrimination reported by the poorest and least-educated average Black individual in any of the samples. These findings suggest that, in 1995-1996 as well as in 2011-2014, as high-achieving Black and White people similarly climb the socioeconomic ladder, Black individuals do it while experiencing high levels of discrimination. These findings confirm evidence showing deteriorated health profiles of Black people (vs. White people) especially in the higher ranks of the labor hierarchy (Borrell et al., Reference Borrell, Kiefe, Diez-Roux, Williams and Gordon-Larsen2013; Dyke et al., Reference Dyke, Vaccarino, Dunbar, Pemu, Gibbons, Quyyumi and Lewis2017). However, this doesn’t mean that White individuals are improving their overall discriminatory experiences. Indeed, our findings shows that White individuals of all educational and income levels manifested higher levels of discrimination in 2011-2014 than about two decades before.
Our findings show, however, that race-income patterns in discrimination in 1995-1996 are not the same as those in 2011-2014. In 1995-1996, a pattern of diminishing returns to income among Black individuals was clear. However, by 2011-2014, Black individuals show similar high levels of discrimination across all income tertiles, with low- and high-income earners showing the highest levels. Black individuals at the bottom of the income distribution reported almost triple the level of overall discrimination in 2011-2014 than almost two decades before whereas those at the top of the income distribution about half (discrimination levels for those in the middle tertile remained practically unchanged). These results imply that, while high-income Black Americans discriminatory experiences are improving, those of the poor are deteriorating. Black individuals went from a situation of diminishing returns to income in 1995-1996 to one of no protective effects of income by 2011-2014. Research on racial disparities in health and in lived experiences of discrimination should be aware that, as the racial gap in overall discrimination is shrinking along the income distribution, it is mostly at the expense of White Americans’ increasing reporting of discrimination at all income tertiles. Our results reflect the stagnation in health gains in the U.S., especially as Black and White Americans are becoming increasingly exposed to lifetime discriminatory practices and experiences in a wide range of life domains.
These patterns in overall discrimination were reflected in our analyses including all types of discrimination (Appendix). The patterns revealed also illustrate the drastically different biopsychosocial experiences of Black and White individuals across a great variety of life domains. In the United States, “success” has a different meaning and life trajectory depending on race (Harris et al., Reference Harris, Cormack, Tobias, Yeh, Talamaivao, Minster and Timutimu2012). At the populational level, race exposes vast heterogeneity in lifetime trajectories—health, discriminatory, socioeconomic—in both relative and absolute terms (Gee and Walsemann, Reference Gee and Walsemann2009). The heterogeneous protective effects of improving two central social determinants of health—education and income—by race, brought to light critical implications for population health, health policy, and the prescription of effective interventions to improve the social determinants of health and the life standards of vulnerable populations (Darling-Hammond, Reference Darling-Hammond1998; Rodriguez et al., Reference Rodriguez, Bae, Geronimus and Bound2021). Patterns are changing fast and, unfortunately, in many instances, deteriorating. Such heterogeneity reminds us that improving the overall health profile of populations is imperative and is best done within a framework of equity and social justice. This is a necessary condition not only for truly effective policy, but also to dismantle entrenched structural racism and classism. The experience of discriminatory treatment is deeply related to social, psychological, and cultural factors, as well as to political actors and institutions, and economic and criminal justice systems that shape the biopsychosocial coping mechanisms employed by individuals to attenuate environmental and institutionally-induced stress (Rodriguez et al., Reference Rodriguez, Umaña-Taylor, Smith and Johnson2009; Rodriguez et al., Reference Rodriguez, Bae, Geronimus and Bound2021).
Our findings also indicated that neither policies nor public interventions should be single-issue implementations: efforts on how to shrink stress-related, discrimination health effects need to evolve into a multi-systems agenda. Some researchers are noting that policy solutions have been historically doomed to be much less effective than intended as they focus too narrowly on solving one problem at a time while leaving other interconnected problems unattended (Dyke et al., Reference Dyke, Vaccarino, Dunbar, Pemu, Gibbons, Quyyumi and Lewis2017). While furthering income and education have the potential to unintendingly entrench racial disparities in life experiences and health, researchers have pointed the growing resistance of low-education Whites who, for political and ideological reasons, oppose the public programs that demonstrably improve their health and socioeconomic wellbeing, even though they situate among the highest beneficiaries of such programs [73].
Policies and interventions to increase income and education should be simultaneously coordinated with anti-racism efforts that reduce discrimination across interconnected contexts of social, economic, and political activity (Williams et al., Reference Williams, Yu, Jackson and Anderson1997) Research reports that students of color frequently experience racial microaggressions that fabricate often-overlooked circumstances of persistent psychosocial disadvantage throughout their academic and personal development (Lee et al., Reference Lee, Peckins, Heinze, Miller, Assari and Zimmerman2018). Interconnected mechanisms to trigger race equity should incorporate as many of the surrounding actors, factors, and organizational settings that correlate with discrimination. This multi-systems policy approach should implement within- and between-institutional layers of the underlying mechanism that institutionalizes racism and discrimination. For example, education policy should reach out and incorporate other key contexts in which discriminatory practices are entrenched—e.g., the labor and housing environments, and their political and social intersections.
Our findings also suggest that it is not enough, for achieving racial health equity, to champion policies and programs that would assist racial minorities to access the educational levels necessary to reach higher-income professional positions. At the moment, for Whites, education and income still operate as buffers that reduce their discrimination experiences; yet, these favorable effects are shrinking in the last few decades. For Black people, education operates as a catalyst while income returns are nonexistent. If anti-racism educational efforts and monitoring mechanisms do not reach out of the labor environment onto the wider social, business, and housing environments, then policies and interventions could harmfully translate into sustained, subtle mechanisms for “inclusionary discrimination” [68]—i.e., a modus operandi by dominant groups in the social hierarchy that “includes” oppressed minorities as they are simultaneously used to discriminate against same-minority members that do not meet the dominant group’s standards for inclusion [69].
Without stretching policy actions through the intersectional web of stigma and social exclusion, as Black Americans gain access to labor and housing privileged settings through higher income and education, they will continue to experience the contradictory expectation of higher levels of discrimination. The goal behind policy implementation is to penetrate the multivariable layers of the causal structure that otherwise would produce and perpetuate racial and SES inequalities. A multi-systems policy approach against enduring discrimination cannot be put in place without tackling the underlying historical, institutional, legal and sociopolitical structures that produce and distribute the social determinants of health (Rodriguez et al., Reference Rodriguez, Bound and Geronimus2014; Rodriguez et al., Reference Rodriguez, Geronimus, Bound and Dorling2015; Williams et al., Reference Williams, Yu, Jackson and Anderson1997). Sociopolitical power is an indispensable resource employed to materialize policy into community benefits. The most notable manifestations of racism shaping racial disparities in health—i.e., residential segregation, mass incarceration, and premature mortality—are therefore functions of the power that different groups in the social hierarchy—with different interests and historical paths—invest in the government and governance (Rodriguez et al., Reference Rodriguez, Geronimus, Bound and Dorling2015).
Some limitations of this study are worth mentioning. First, the subsample of Black individuals in our samples is relatively small thus increasing imprecision in the estimates. However, our statistical results were highly consistent both using an overall index of discrimination and across a great variety of types of discrimination (Appendix), confirming patterns in studies of lifetime discrimination with larger subsamples of Black individuals. Second, the Black subsamples included in both MIDUS 1 and MIDUS Refresher are not a nationally representative samples of Black Americans, reminding us that generalization of results should be made with caution. Yet, our application of a compound weighting strategy improved comparability between White and Black participants while allowing us to compare our results in time. Third, our samples show a higher socioeconomic status than the overall U.S. White and Black populations; yet, both Whites and Black Americans in the samples manifest great variation across many of the key indicators we used in our analyses thus including, nevertheless, many individuals of low socioeconomic status. Additionally, we also applied a propensity scores matching technique that effectively enhanced the statistical comparability between Black and White participants accounting for the heterogeneity of our samples (Clouston and Link, Reference Clouston and Link2021).
Taken together, our findings add to the growing record on racial disparities in perceived lifetime discriminatory experiences, especially those that are known to bring long-term health effects. We identified substantive and robust diminishing returns of education across time and a continuing income discrimination disadvantage among Black individuals. As Black individuals increase their education, the more they perceive being discriminated. As Black individuals climb the educational ladder, structural oppression increases thus injecting the negative health repercussions of high-coping effort. Also, the U.S. seems to be in track toward a stagnation of public health gains. From 1995-1996 to 2011-2014, Black individuals manifested worse discriminatory experiences than Whites across educational and income levels while White individuals of all levels of education and income reported increased levels of discriminatory experiences. Racial disparities in discrimination are shrinking at the expense of an overall deterioration of discriminatory experiences, especially among White individuals. Necessary policies and interventions to increase education and income and, therefore, improve population health, paradoxically threatens to deflect our attention from historically embedded factors that differentially distribute or obstruct the benefits of the social determinants of health. The dismantling of lifetime discrimination across all affairs and institutions of society suggests a multi-systems policy-making approach—one that avoids overemphasizing single problems and, instead, directs our efforts toward the interconnected web of causal factors that entrench biopsychosocial differences by socioeconomic status and race.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0021932022000360.
